Causal Modeling of Policy Interventions from Treatment-Outcome Sequences

This repository contains the public code to reproduce experiments and figures of the paper "Causal Modeling of Policy Interventions From Treatment-Outcome Sequences" by Hızlı et al., ICML 2023.

Citation

If you build on top of this code, please cite

@InProceedings{hizli23a,
  title = 	 {Causal Modeling of Policy Interventions from Treatment--Outcome Sequences},
  author =       {H{\i}zl{\i}, \c{C}a\u{g}lar and John, S. T. and Juuti, Anne Tuulikki and Saarinen, Tuure Tapani and Pietil\"{a}inen, Kirsi Hannele and Marttinen, Pekka},
  booktitle = 	 {Proceedings of the 40th International Conference on Machine Learning},
  pages = 	 {13050--13084},
  year = 	 {2023},
  volume = 	 {202},
  series = 	 {Proceedings of Machine Learning Research},
  pdf = 	 {https://proceedings.mlr.press/v202/hizli23a/hizli23a.pdf},
  url = 	 {https://proceedings.mlr.press/v202/hizli23a.html},
}

Development environment and dependencies

Python

The code has been developed and tested with Python 3.9 under MacOS X. Under other operating systems, you have to adjust the tensorflow package in requirements-with-deps.txt. After activating a python environment, run

pip install -r requirements-with-deps.txt
pip install -r requirements-without-deps.txt

R

The baseline of Hua et al. (2021) requires R. The code has been tested both with R v4.2.0 and v3.6.3. It requires the locally provided doct package and reticulate from CRAN, which can be installed with the following code:

install.packages("remote")
install_local("models/benchmarks/hua/doct/doct_1.0.tar.gz")
install.packages("reticulate")

Reproducing experiments

To reproduce the experiments on real-world data (section 7.1), run

PYTHONPATH=. bash run_experiment_rw.sh

This will recreate

Figure 4 in out/figures/figure3/train_fit_d1_p8_fig3.pdf
Figure 7 in out/figures/figure7/train_fit_d1_p8_fig7.pdf
Figure 8 in out/sampler/outcome.p3,8,12/figures/f_all_fit_id{0,1,2}_vTrue.pdf
Figure 9 in out/sampler/action_pred/compare_{ga,go,mark_intensity}.pdf
Figure 10 in out/sampler/outcome.p3,8,12/figures/compare_{fb,ft_m3}.pdf
Figure 11:
- (a) in figures.figure11/fig11_real_data_p3_d0.pdf
- (b) in figures.figure11/fig11_real_data_p8_d1.pdf
- (c) in figures.figure11/fig11_synth_data_p3_s12,15,27.pdf
- (d) in figures.figure11/fig11_synth_data_p8_s1,4,25.pdf

To reproduce the experiments on semi-synthetic data (section 7.2), run

PYTHONPATH=. bash run_experiment_synth.sh

(this will fit to real-world data to run the sampler to generate the semi-synthetic data).